David Miller wrote: [Sat Oct 04 2014, 07:26:30PM EDT] > > Third time is the charm, right? :-) Let's hope so... > > Changes since v2: > > 1) Get rid of verbose kernel log messages during vmemmap construction. > > 2) Add patch #9 to get rid of superfluous page table blocks in the > kernel image which can be allocated dynamically, and at the same > time increase MAX_BANKS to hopefully accomodate guests with very > fragmented real memory layouts. > > 3) Fix bug in KERN_PGTABLE_WALK that would corrupt the huge page TTEs > of vmemmap mappings. > > 4) Do some type checking of the page table pointer objects passed into > pgd_populate() and pud_populate() by creating inline function > helpers. The reason we can't make pgd_populate() and > pud_populate() inline is that "struct mm_struct" is sometimes not > fully defined yet in contexts where this header is included. > > Anyways, this found a bug (albeit in the end harmless) in the new > implementation of vmemmap_populate() in that we were passing a > "pgd_t *" instead of a "pud_t *" into the third argument of > pgd_populate. > > As usual, let me know how it goes. > > Signed-off-by: David S. Miller <davem@xxxxxxxxxxxxx> Remote M7-4 up with spec-jbb and hugepages. Local T4-2 up with DEBUG_PAGEALLOC. Remote T5-8 being severely thrashed with several different loads concurrently. Local T5-2 was under severe load with THP "always" until my "QuietComfort 3" headphones declared magnificent/enough :) Seriously, I believe we are in good condition. There will be issues but this is always the challenge. Should you desire a boot log, please feel free to request. I've consumed nearly all boot logs with scrutiny. thanx! Acked-by: Bob Picco <bob.picco@xxxxxxxxxx> -- To unsubscribe from this list: send the line "unsubscribe sparclinux" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html